Brandon Gallagher
Book a call
AI Platform Engineer · High-Consequence Systems

Brandon Gallagher

Deterministic agentic infrastructure—built at extreme velocity

I own compliant landing zones and control planes end-to-end: IaC, containment, Kubernetes delivery, and production AI harnesses for high-consequence financial systems. No ticket theater. No manual ops debt.

Book a call Download resume

Platforms shipped for

  • Scotiabank
  • CIBC
  • Haventree Bank
  • Interac
  • RBC
  • Deloitte
Operating Rules

Engineering philosophy—non-negotiable.

Opinionated standards for shipping agentic infrastructure under compliance pressure without trading velocity for rigor.

  1. Automate the pattern, not the ticket

    Recurring ops work is a systems failure. Resolve classes of problems with workflows and control planes—never grow a queue.

  2. Security lives in IaC

    PCI-DSS, SOC 2, and NIST are encoded in Terraform and pipelines. If it isn’t in the graph, it doesn’t ship.

  3. Zero tolerance for manual operational debt

    Runbooks that require heroics are unfinished design. Self-service account vending, deterministic deploy paths, no snowflake clusters.

  4. Deterministic boundaries for agentic systems

    Containment, policy, and observability are first-class architecture—not a bolt-on after the demo works.

  5. Own the architecture end-to-end

    Trade-offs are explicit and owned. Hosted vs open-source, managed vs DIY—chosen for deploy velocity and blast-radius control.

Active Now

Shipping high-consequence architecture now.

Current build where agent containment, policy gates, and observability are load-bearing.

RAVL
Live Deployment

AI Platform Engineer

High-performance agentic harness

Production control plane for multi-agent workloads: strict containment, policy gates, and observable tool use—designed for enterprise blast-radius limits.

Active Stack

  • LangGraph
  • Kubernetes
  • AgentGateway
  • GCP
Platforms Shipped

Proof / platforms shipped

Hard metrics and architectural trade-offs from a tech leader across payments, platforms, and national programs

Nezz
Impact
Programmable payments machine in production—deterministic state machines, multi-party settlement paths, and audit-ready event trails (CTO & Co-Founder)
Architecture
Settlement orchestration as state machines—not ticket-driven payment ops; backed by operators from BMO, Amex, Mastercard, and MIT Innovation Labs
Trade-off
Institutional settlement semantics first—audit trails and multi-party paths before feature breadth
  • TypeScript
  • Terraform
  • Kubernetes
  • Azure
Scotiabank
Impact
Shipped an enterprise Agentic Control Plane with the AI Consortium—deterministic containment for production LLM workloads
Architecture
Hosted control plane over open-source DIY—governance and deploy velocity beat flexibility at bank scale
Trade-off
Vendor deploy-readiness is an architecture constraint—fail early or integration becomes the critical path
  • Google Cloud
  • Kubernetes
  • LangGraph
  • AgentGateway
CIBC
Impact
160+ repos migrated to GitHub Actions—1,200+ engineering hours/year recovered via AI auto-remediation
Architecture
Org-wide shared workflows over per-repo pipelines—compliance scripts mandatory, not optional
Trade-off
Align Node runtimes before enterprise rollout—version drift blocks adoption overnight
  • GitHub Actions
  • GitHub
  • OpenShift
  • Node.js
Haventree Bank
Impact
Tier 1 CDIC fintech greenfield → production on Azure under PCI-DSS, SOC 2, and NIST
Architecture
Container Apps + Terraform Stacks; Azure Firewall over Palo Alto to shrink managed ops surface
Trade-off
Bake compliance into IaC from day one—auditors read the graph, not a wiki
  • Azure
  • Terraform
  • Kubernetes
  • Docker
Interac
Impact
AWS Control Tower landing zone in Terraform for e-Transfer—governed multi-account for 30M+ Canadians
Architecture
Control Tower + Terraform over ad-hoc account provisioning—zero-friction account vending
Trade-off
Landing zones only scale when account vending is self-service—not ticket-driven
  • AWS
  • Terraform
  • HashiCorp
  • Kubernetes
RBC
Impact
Automated 5,000+ IAM Jira requests on OpenShift—cleared a recurring ops bottleneck with zero added headcount
Architecture
Node.js microservice automation over manual ticket queues—pattern resolution at scale
Trade-off
High-volume IAM is a workflow problem—automate the class, not the individual ticket
  • OpenShift
  • Node.js
  • Jira
Deloitte
Impact
Canada Post national track-and-trace—real-time stream integrity for 35M+ Canadians
Architecture
Kafka stream-validation microservices over batch-only reconciliation
Trade-off
Catch bad data at ingress—every downstream consumer pays for upstream gaps
  • Kafka
  • TypeScript
  • Node.js
  • Kubernetes
Capability

Control planes, landing zones, and Kubernetes—owned end-to-end.

The stack I ship with when the blast radius is real: deterministic agent boundaries, compliance encoded in IaC, and production clusters that don’t require heroics to keep alive.

  • Agentic Control Planes
    Harnesses with strict containment, policy gates, and automated governance—production agent systems without hoping the demo holds.
    • LangGraph
    • agentgateway
    • vLLM
  • Compliant landing zones
    Hub-and-spoke foundations where Terraform + Sentinel encode PCI-DSS, SOC 2, and NIST—account vending as product, not tickets.
    • Azure
    • AWS
    • GCP
    • Terraform
    • Terraform Cloud
    • Sentinel
    • AWS Control Tower
    • PCI-DSS
    • SOC 2
    • NIST 800-53
  • Kubernetes at scale
    Production AKS / OpenShift delivery with Helm, RBAC, and geo-redundancy—proven on financial platforms serving 30M+ users.
    • Kubernetes
    • AKS
    • Helm
    • Docker
    • Azure Container Apps
    • GitHub Actions
    • GitHub Enterprise
    • Azure DevOps
Get In Touch

Ready to ship. Let’s talk.

Direct email or thirty minutes on the calendar—no forms, no gatekeeping.

brandonegallagher@gmail.com Book a call Resume